Following evolution of bacterial antibiotic resistance in real time

Following evolution of bacterial antibiotic resistance in real time
Adam Z Rosenthal & Michael B Elowitz
A new study reports the development of the ‘morbidostat', a device that allows for continuous culture of bacteria under a constant drug selection pressure using computer feedback control of antibiotic concentration. This device, together with bacterial whole-genome sequencing, allowed the authors to follow the evolution of resistance-conferring mutations in Escherichia coli populations in real time, providing support for deterministic evolution of resistance in some situations.
Microorganisms can and do evolve resistance norfloxacin. This study identified an altruistic The morbidostat al ows automated evolution to antibiotics. In recent decades, the spread mechanism for antibiotic resistance whereby experiments over extended timescales of weeks of bacterial drug resistance has been acceler- resistance in a smal fraction of cel s could sup- or longer. Furthermore, because cel s are con- ated by the widespread use of antibiotics in port a larger (drug-sensitive) population.
tinual y expel ed from the device, it is possible agriculture and in healthcare. This trend to sample cel s in a nearly continuous manner has raised fundamental and urgent ques- Morbidostat
during the evolutionary experiment and to tions about the dynamics and genetic basis of In such laboratory evolution studies, it has fol ow their genetic changes by whole-genome bacterial resistance. What specific mutations been technical y chal enging to analyze evo- sequencing of cel s captured at each time point. arise in response to low drug concentrations, lutionary adaptation to continuous selection Final y, by maintaining 15 replicate cultures in and how do additional mutations strengthen pressure from antibiotics under constant con- parallel, Toprak et al.1 were also able to explore resistance at higher drug concentrations? Does ditions. The problem is that as cel s become the reproducibility of mutational trajectories resistance result from mutations in just one resistant to an antibiotic, its selection pressure during paral el evolution.
or a few genes, or is it spread more broadly is reduced or eliminated. To compensate, one across the genome? Final y, to what extent are must continual y increase the antibiotic con- Evolution of resistance
the specific mutations that confer antibiotic centration in proportion to the increased resis- Toprak et al.1 used the morbidostat to examine resistance and the temporal sequence in which tance in the culture4.
the evolution of drug resistance in E. coli pop- they arise predictable? Addressing these ques- To address this chal enge, Toprak et al.1 ulations grown for up to 25 days under con- tions requires the ability to map mutations at have now developed a sophisticated computer- tinuous chal enge with one of three different a genome-wide scale and to do so over time control ed microbial cultivation system that can antibiotics: trimethoprim, chloramphenicol or during a continuous evolutionary process. On maintain a constant antibiotic selection pres- doxycycline. They performed whole-genome page 101 of this issue, Roy Kishony and col- sure. Their device, which they name the ‘mor- sequencing of strains sampled over time, iden- leagues1 achieve such real-time measurements bidostat', is a modern version of the turbidostat. tifying specific resistance-conferring muta- by combining whole-genome sequencing with In both devices, fresh medium enters a vessel at tions and reproducible temporal sequences of a new, powerful feedback system for continu- the same rate that culture medium containing their appearance.
ous cultivation of bacteria under constant cells is expelled, and this rate can be controlled In the case of trimethoprim, they found that selection pressure. This technical advance in in response to continuous measurements of cell the majority of mutations for resistance mapped laboratory-based evolution is beginning to density (Fig. 1)5. The morbidostat inserts a to DHFR, the gene encoding the known target
answer many of these questions.
computer program into this feedback loop of this antibiotic, dihydrofolate reductase. Microbial evolution studies have been revo- and adds additional control over antibiotic Moreover, when the stepwise accumulation lutionized by recent applications of whole- concentration. It thereby implements a simple of DHFR mutations was tracked by sequenc- genome sequencing. For example, Lenski and feedback loop in which faster cel growth leads ing whole genomes over time, mutations col eagues in their groundbreaking evolution to more antibiotic in the culture, which in turn were found to occur in a particular temporal experiment identified the specific genetic leads to slower growth, keeping cel densities sequence in paral el cultures, suggesting that changes in E. coli that occurred over the course constant over long periods of time despite evo- evolutionary adaptation can be deterministic6. of 50,000 generations in the laboratory2,3. This lutionary adaptation.
This result is broadly consistent with previ- work showed the genetic effects of evolutionary The morbidostat thus acts like a virtual ous predictions from Hartl and colleagues7, adaptation to a repeated environmental regi- treadmill whose speed (antibiotic concentra- who synthesized a set of combinatorial men (glucose limitation) in paral el cultures. tion) is determined by cel growth rate. As cel s mutations in the gene encoding B-lactamase, More recently, Col ins and col eagues4 used develop resistance to the antibiotic and begin which is involved in antibiotic resistance. whole-genome sequencing to analyze the evo- to grow faster (running toward the front of These studies, which were limited to a small lutionary adaptation of E. coli to the antibiotic the treadmill), the morbidostat, like a tough fraction of possible mutations in just one coach, compensates by increasing antibiotic gene, suggested that some mutational paths Adam Z. Rosenthal and Michael B. Elowitz
concentration (treadmil speed) just enough could generate monotonical y increasing fit- are at the Howard Hughes Medical Institute,
to keep cel s at the original growth rate. ness for the cell, although most could not. Division of Biology, California Institute of
Cel s that can't keep up die, or are swept off Toprak et al. now provide confirmation that Technology, Pasadena, California, USA.
the back of the treadmil , as the most resistant relatively deterministic mutation sequences e-mail: [email protected]
proliferate faster.
are indeed possible, even when evolution is NATURE GENETICS VOLUME 44 NUMBER 1 JANUARY 2012

In a complementary approach to that taken by Toprak et al., another group recently developed a microfluidic culture device that addresses issues of spatial heterogeneity in bac- terial growth8. Natural environments may have strong spatial gradients of antibiotics, which could affect the evolution of resistance9. To mimic such conditions, Zhang et al.8 created a hexagonal two-dimensional array of chambers connected by narrow channels and applied a gradient of the antibiotic ciprofloxacin to the device. They followed the growth and move- ment of a population of cel s that were initial y sensitive and sequenced the whole genomes of resistant mutants.
Increasing antibiotic resistance Beyond the laboratory
To what extent does this laboratory example Conventional selection of paral el evolution apply to real-life situ- ations? The spread of an epidemic provides an opportunity to track parallel bacterial evolution, as the same strain infects multiple individuals in parallel. Lieberman et al.10 pro- vide an example of this using whole-genome sequencing of frozen clinical Burkholderia Antibiotic concentration dolosa isolates taken from patients with cystic fibrosis followed in a single hospital over a 16-year period. Strikingly, similar bacterial mutations were seen to arise independently Morbidostat selection in different patients, as the pathogen evolved through within-host evolution and transmis- sion between patients. A complementary study by Sebastien Gagneux and col eagues11, on page 106 of this issue, analyzed the occur- rence of rifampicin resistance during tuber- culosis infections. This work identified a Antibiotic concentration two-step process in which initial mutations in rpoB, the B subunit of the RNA polymerase, Figure 1 The morbidostat applies a continuous selective pressure to the bacterial population. conferred resistance and were fol owed by (a) The morbidostat maintains a constant growth rate through feedback control of the mixture of medium (yellow) and medium with antibiotic (red) inflow. As bacteria acquire high levels of resistance to the the acquisition of compensatory mutations antibiotic, they are able to tolerate higher antibiotic concentrations (depicted in red) while maintaining in other polymerase subunits that al owed for a constant growth rate. Thus, antibiotic concentrations increase progressively over time (changing increased fitness.
color). (b,c) Evolution of antibiotic resistance in conventional and morbidostat models of antibiotic Together, these laboratory microbial evo- resistance. Bacterial growth rate (green line) is shown relative to increasing antibiotic concentration (red lution and epidemiological studies offer key line). Black arrows indicate resistance-conferring mutations. In conventional antibiotic selection (b), insights into the evolution of bacterial drug selective pressure is applied by adding a sudden antibiotic challenge to reduce bacterial growth rate. Resistance-conferring mutations emerge and allow the population to regain growth. Under selection resistance. These studies have identified in the morbidostat (c), the bacterial growth rate is held constant as antibiotic concentrations are not only new resistance-conferring muta- automatically adjusted using computer feedback to compensate for acquired mutations that affect tions, but sometimes even the order in which the growth rate.
they occur. Furthermore, these results sup- port the view that microbial evolution can unconstrained rather than artificial y limited associated with a more general form of multi- be deterministic and, therefore, potentially to a single locus.
drug resistance. Nevertheless, most of the predictable in some scenarios. The interface Such deterministic genetic trajecto- mutations clustered into a relatively small between bacterial laboratory and natural evo- ries are not universal, however. When the number of operons, suggesting that there may lution studies wil provide fertile ground for authors used the morbidostat to analyze be a limited number of ways for the cel to cir- developing new concepts to understand the evolutionary responses to two other drugs— cumvent a given antibiotic. However, in any processes of resistance and evolution and new chloramphenicol and doxycycline—they particular culture, the cells did not exhaust technologies to apply to increasingly urgent observed qualitatively different behaviors. In the full evolutionary potential provided by real-life problems.
response to these antibiotics, they identified the complete set of advantageous mutations, resistance-conferring mutations in a larger despite the increase of antibiotic concentra- COMPETING FINANCIAL INTERESTS
number of gene targets, many of which are tion by over three orders of magnitude.
The authors declare no competing financial interests.
VOLUME 44 NUMBER 1 JANUARY 2012 NATURE GENETICS

1. Toprak, E. et al. Nat. Genet. 44, 101–105 (2012).
5. Bryson, V. & Szybalski, W. Science 116, 45–51 8. Zhang, Q. et al. Science 333, 1764–1767 2. Barrick, J.E. et al. Nature 461, 1243–1247 6. Gould, S.J. Wonderful Life: The Burgess Shale 9. Hermsen, R. & Hwa, T. Phys. Rev. Lett. 105, 248104 3 Woods, R.J. et al. Science 331, 1433–1436 and the Nature of History 1st edn. (W.W. Norton, 10. Lieberman, T.D. et al. Nat. Genet. 43, 1275–1280 4. Lee, H.H., Molla, M.N., Cantor, C.R. & Collins, J.J. 7. Weinreich, D.M., Delaney, N.F., Depressor, M.A. & Nature 467, 82–85 (2010).
Hartl, D.L. Science 312, 111–114 (2006).
11. Comas, I. et al. Nat. Genet. 44, 106–110 (2012).
Dnmt3a silences hematopoietic stem cell self-renewal
Jennifer J Trowbridge & Stuart H Orkin
DNA methylation is an epigenetic mark stably directing gene expression throughout development. A new study uncovers a role for the DNA methyltransferase Dnmt3a in silencing self-renewal genes in hematopoietic stem cells (HSCs) to permit efficient hematopoietic differentiation.
Proper orchestration of epigenetic patterns stem cel turnover. In the secondary transplant Dnmt3a in repression of the stem cel program, and transcription factors is critical to ensure recipients, there was a striking 200-fold expan- which it carries out by silencing HSC genes and hematopoietic homeostasis. The recent dis- sion in phenotypic HSCs, which was sustained permitting differentiation, presumably via covery of mutations in the DNA methyl- through four rounds of serial transplantation. regulation of DNA methylation.
transferase gene DNMT3A in human acute These HSCs did not exhibit increased prolif- myeloid leukemia (AML)1 and myelodysplastic eration or resistance to apoptosis, suggesting Direct and indirect mechanisms
syndrome (MDS)2 highlights the importance that an intrinsic self-renewal program was These provocative findings raise intrigu- of DNA methylation in hematopoietic malig- activated. Consistent with this hypothesis, ing questions about the role of Dnmt3a in nancies. However, the underlying mechanisms Dnmt3a-null HSCs showed increased expres- hematopoiesis. The Dnmt3a-null HSC pheno- by which Dnmt3a regulates hematopoiesis and sion of genes implicated in HSC self-renewal type, revealed by serial transplantation, HSC function remained unclear. On page 23 and multipotency, including Runx1 and Gata3 was likely overlooked in the previous report6 of this issue3, Margaret Goodel , Wei Li and col- (Fig. 1a). Despite the remarkable expansion of or perhaps was a result of the different Cre
leagues show that Dnmt3a is critical for silenc- Dnmt3a-null HSCs, their contribution to over- drivers used in these studies. The capacity of ing HSC self-renewal genes, thereby enabling al blood production did not proportional y Dnmt3a-nul HSCs to function normal y in efficient hematopoietic differentiation.
increase, suggesting a differentiation defect in primary transplants cal s into question the these cel s. Dnmt3a-nul HSCs generated the mechanism underlying the dramatic expan- New role for Dnmt3a
ful complement of hematopoietic cel s, with sion of HSCs seen in subsequent transplants. DNA methylation is established and main- a bias toward B-cel differentiation, and their Because the number of HSCs residing in the tained by three DNA methyltransferase multi-lineage differentiation capacity declined bone marrow is limited by available niche space enzymes: Dnmt1, Dnmt3a and Dnmt3b. with successive rounds of transplantation. The within the microenvironment7, the expansion During hematopoiesis, Dnmt1 is required for authors conclude that conditional knockout of of Dnmt3a-null HSCs is surprising, and resi- both HSC self-renewal and differentiation4,5, Dnmt3a impedes hematopoietic differentiation dency of HSCs outside the bone marrow niche whereas Dnmt3a has been reported to be while causing the HSC population to expand warrants further examination. With respect to dispensable in HSCs6. The recent discovery in the bone marrow.
hematopoietic differentiation, the bias toward of DNMT3A mutations in AML1 and MDS2 To uncover the molecular mechanism B-cel differentiation cal s for elucidation of led Challen et al.3 to re-evaluate the role underlying the phenotype of Dnmt3a-null the precise phenotype of Dnmt3a-null lym- of Dnmt3a in HSCs by using a conditional HSCs, Challen et al.3 profiled changes in phoid cel s. As Dnmt1 is necessary to promote knockout mouse model coupled with serial global gene expression and DNA methylation. differentiation into lymphoid progeny4, such transplantation. In primary transplant recipi- Counterintuitively, abundant DNA hyper- elucidation wil provide better insight into the ents, Dnmt3a-nul HSCs contributed normal y methylation was observed along with regions cel context–dependent role of Dnmt3a and the to hematopoiesis, as reported in the previous of DNA hypomethylation in the Dnmt3a-null distinct functions of DNA methyltransferases study6. Challen et al.3 then went a step further HSCs, and the correlation between changes in in hematopoietic differentiation.
by isolating Dnmt3a-null HSCs from the pri- DNA methylation and gene expression was The poor correlation between changes in mary transplant recipients and transplanting weak. In contrast, a net reduction in global DNA methylation patterns and differential them into secondary recipient mice to force DNA methylation was associated with sus- gene expression in Dnmt3a-null HSCs is sur- tained expression of stem cel –specific genes in prising. Because the profiling experiments were Jennifer J. Trowbridge and Stuart H. Orkin
Dnmt3a-nul B cel s. These results highlight the performed at the population level, the findings are at the Department of Pediatric Oncology,
context-dependent consequences of Dnmt3a may be confounded by the heterogeneity of Dana-Farber Cancer Institute and Division
loss. The impaired differentiation capacity of HSCs. The exact relationship between Dnmt3a- of Hematology/Oncology, Children's Hospital
Dnmt3a-nul HSCs was partial y restored in induced methylation and transcription wil only Boston, Boston, Massachusetts, USA.
rescue experiments, indicating that many of be revealed with future technology allowing e-mail: [email protected] or the functional changes are reversible. Taken evaluation of such changes at the single-cell
together, these findings reveal a major role for level. The authors attribute the poor correlation NATURE GENETICS VOLUME 44 NUMBER 1 JANUARY 2012

Source: http://www.elowitz.caltech.edu/publications/Evolution%20in%20Real%20Time.pdf